Which editor fixes what
Say the macro keeps picking Inspire when you’d rather it took Melody. You could raise Melody’s weight, write a rule, paint a zone, or move a curve, and all four would change that pick - but only one of them changes just that pick and leaves the rest of your profile alone. This page matches the decision you saw to the editor that owns it.
First, check the field pattern
Section titled “First, check the field pattern”The most common mistake happens before AI Studio is even open: tuning AI Gather settings or a profile while the field’s gather pattern is still set to Kettle or another fixed pattern. AI Gather only runs on fields whose pattern is set to AI Gather - until that’s switched, nothing you change in a profile matters, and “AI Gather isn’t working” reports are usually this. Fix it on Set up AI Gather first.
Read Trace before you touch anything
Section titled “Read Trace before you touch anything”When a pick looks wrong, expand Trace first. It names the stage that
actually moved the score, which tells you which editor to open. Open
Tools → Simulator, load the scenario closest to the moment you saw (or step
to that moment in Test Bench), and click Trace on the winner. The biggest
line in the breakdown is your answer. If it’s the base weight line, the class
itself is too strong or too weak. If it’s a rule you wrote, the rule is the
problem. If it’s × walk distance or another Built-in row, the curve is.
The editors, by the question they answer
Section titled “The editors, by the question they answer”| What you saw | Open | Why it’s the right one |
|---|---|---|
| Melody loses in every field, in every situation | Token Priorities | Base weight changes where that class starts, everywhere. |
| Melody only loses while its buff is low | Token Rules | An adjust or Gate rule changes one class in one situation. |
| Melody loses only in a corner blocked by props | Field Zones | A zone changes scores based on where your character is standing. |
| Every class walks too far for weak tokens | Tuning | A curve changes how much distance, crowding, or expiry matters to every token. |
| A rare token must beat everything else when it appears | High Value | High Value tokens use their own tier, not a huge normal weight. |
| The macro abandons a return or a token it had almost reached | Interrupt rules | An Interrupt rule decides whether a new candidate may replace what’s already in progress. |
| Only long-lived duped tokens need different handling | Duped | Duped tokens are scored on their own, apart from ordinary tokens. |
Adding a rule or increasing a token’s weight answers “which class should I prefer?” - a zone answers “where on the field should that preference change?” Raising Melody’s weight increases the score of every Melody token. Painting an Avoid zone over a corner blocked by props only changes scores while your character is in that place. That place-based problem is what only zones solve.
A Token Rule answers “this class, in this situation, should be worth more” - a tuning curve answers “how much should distance, crowding, or expiry matter to everything?” Raising Melody’s weight while its buff runs low only affects Melody. Flattening the walk distance curve changes what every token on the field is worth from where you’re standing. That shared shape is what only curves fix.
High Value is for tokens that must be collected at all cost, where missing the collection would be severely detrimental to honeymaking. By default that includes rare tokens such as Bear Morph and Festive Gift, which can only be made by a single bee every so often. When several High Value tokens are on the field, the one highest in the list is collected first, and ordinary rule adjustments can’t accidentally steal a token you’ve marked High Value.
Actions are a different kind of decision. An action such as Return to Sprinkler, a camera rotation, or a pattern has no distance, no cluster, and no expiry, so the curves that shape a token’s score mean nothing for it. Its score comes entirely from the rules you write under Configure → Actions. Actions and action rules covers them.
The mistake to avoid
Section titled “The mistake to avoid”Inside AI Studio, the mistake is reaching for weight when the problem sits at a different stage of the chain. Stacking add rules until far junk outscores everything defeats the walk distance curve you already have. Stacking gates and expecting the newest one to win ignores that Block always beats Allow. Painting a zone to fix a class problem, or writing ten rules to fix a field problem, both fight the design instead of using it. The chain gives you an editor for each kind of problem - the mistake is always using the loudest one.
After you’ve made the change, test it in Simulator before you make another. Change one thing, then check it is the loop.